The effects of weak versus strong relational judgments on response bias in Two-Alternative-Forced-Choice recognition: Is the test criterion-free?

It is widely believed that a Two-Alternative-Forced-Choice (2AFC) in an old/new recognition memory test is made by comparing the two items and choosing the item with the higher strength. For this reason, it is considered to be criterion-free by some researchers. We found evidence that subjects probabilistically compromised the comparison by choosing the left item when they recognized it as old. Using both normal test pairs (comprised of one new and one old item) and two types of null pairs (comprised of both-new or both-old items), we found that a left-biased choice was coupled with higher hit and false alarm rates and a shorter left than right-choice RT for the normal pairs, consistent with the hypothesis of a bias for making a choice on the basis of a left individual-item recognition. For the null pairs, RT was much longer for the both-new than for the both-old pairs, providing additional evidence for basing decision on an individual-item's absolute, rather than a relative, familiarity. Additionally, subjects gave higher confidence ratings to choices for the both-old than the normal and both-new pairs, again suggesting that their decision was based on absolute familiarity of the items. The results were found to be not due to a fast-response instruction. A comparative judgment experiment in which subjects chose the item higher or lower in an attribute magnitude did not show the response side bias and RT asymmetry. The presence of bias in the former, and the absence of it in the latter can be explained by a weak versus strong relational judgment in the former and the latter type of 2AFC, respectively. We discuss the implications these findings have for the use of the 2AFC as a method for testing recognition memory.